Concept-based semi-automatic classification of drugs.

Area of Science:

• Pharmacology
• Bioinformatics
• Computational Chemistry

Background:

• The World Health Organization's Anatomical Therapeutic Chemical (ATC) system is a global standard for drug classification and drug repurposing.
• The current ATC system has limitations, omitting several major global drugs, hindering comprehensive drug repurposing research.

Purpose of the Study:

• To develop and evaluate a novel classification approach for unclassified drugs using information extraction and machine learning.
• To enhance the existing Anatomical Therapeutic Chemical (ATC) system by assigning putative ATC class labels to previously unclassified drugs.

Main Methods:

• An information extraction (IE) framework was developed to extract drug-related terms (chemical, pharmacological, therapeutic, systemic effects) from scientific literature.
• Extracted terms were utilized as features in a machine learning (ML) framework to predict ATC class labels for unclassified drugs.
• The system was validated on cardiovascular system drugs, employing 100-fold bootstrapping and an independent test set.

Main Results:

• The developed concept-based classification system achieved a predictive accuracy of 89.47% with bootstrapping and 77.12% on an independent test set.
• The IE and ML-based approach demonstrated superior performance compared to state-of-the-art methods relying on chemical structure properties.

Conclusions:

• The proposed IE and ML approach offers a robust method for classifying unclassified drugs and expanding the utility of the ATC system.
• This advancement has significant implications for drug repurposing research by providing a more complete drug classification framework.